Hi! I am Kabir, a PhD student at Paul G. Allen School of Computer Science, University of Washington (UW), where I am fortunate to be advised by Prof. Yulia Tsvetkov. Before joining UW, I spent two wonderful years at Microsoft Research India as a Pre-doctoral Research Fellow, where I worked with Dr. Sunayana Sitaram, Dr. Monojit Choudhury, and Dr. Navin Goyal. For what feels like eons ago now, I did my undergraduate studies at BITS Pilani, India, where I graduated with a B.E (hons.) in Chemical Engineering (story 🍿 for some other day).
I am broadly interested in Natural Language Processing (NLP), mainly towards understanding and improving the capabilities of language models. Some specific areas that I am currently working on or have looked at in the past include:
🔭 Science of Language Models. I am fascinated towards gaining an in-depth understanding of how language models work, their capabilities, and limitations. Some of my work in this space has included studying inductive biases that influence hierarchical generalization in transformers (TACL), understanding in-context learning through a Bayesian perspective (ICLR 2024), and analyzing capabilities and limitations of transformers and RNNs by studying them on Formal languages (EMNLP 2020, COLING 2020).
🌎 Multilingual NLP. How can we build linguistically fair NLP models that can serve a wide range of languages of the world including low-resource languages and dialects with little to no available data? In my past work I have looked at benchmarking propriety LLMs on 70 typologically diverse languages on multiple NLP datasets (EMNLP 2023) highlighting the limitations of current models in multilingual settings, studying cost-performance trade-offs of training multilingual models with different sources and qualities of data (NAACL 2022), and studying cross lingual transfer in pre-trained multilingual models (ACL 2022, EMNLP 2022).
For a full list of my publications you can have a look here. Please feel free to reach out to me over email if you have any questions about my research.
2024-12-02: Our work on hierarchical generalization is now accepted in TACL - Preprint here.
2024-07-23: Attending ICML 2024 at Vienna for the spotlight presentation our work on hierarchical generalization at the Mechanistic Interpretability Workshop.
2024-04-25: Preprint for our work on hierarchical generalization out! - Check the paper here.
2024-01-16: Our work on in-context learning through a Bayesian perspective got accepted at ICLR 2024 for poster presentation. - Check the paper here.
2023-09-26: Started my PhD in Computer Science at University of Washington 🌸.
Courses
Talks and Tutorials
Tutorial on Everything you need to know about Multilingual LLMs: Towards fair, performant and reliable models for languages of the world at ACL 2023. Tutorial slides available here.
Talk on multilingual evaluation of large language models for the Microsoft Africa Research Institute’s research seminar. Recording available here.
I am also happy providing mentorship to students looking to start their research journey in NLP. Feel free to reach out to me over my email.
Theme by Ankit Sultana. Yi Yi illustration at the bottom right by Anna Vignet.